Fusing Local and Regional Datasets to Develop a Composite Land Cover Product Across High Latitudes

Abstract. Rapid warming across the Arctic is the primary driver of widespread permafrost thaw, with far-reaching consequences for local ecosystem resilience, the regional carbon budget, and the global climate system. Because permafrost characteristics and vulnerability are tightly linked to land cover, particularly vegetation type and surface properties, understanding these dynamics requires accurate and detailed land cover information. Spatial variation in vegetation cover influences energy balance, snow insulation, and soil moisture, factors that directly affect permafrost stability. Consequently, high-resolution land cover products are essential for assessing the ecological impacts of permafrost thaw and for improving the representation of permafrost-related processes in predictive models. However, many global land cover datasets fail to capture the spatial heterogeneity and fine-scale ecological features that influence permafrost dynamics, while more detailed regional products often lack coverage across broader, continental extents. This gap presents a challenge for large-scale assessments of permafrost vulnerability under accelerating climate change.

To create a spatially cohesive land cover map that accurately represents the distribution of ecosystems across the Arctic-Boreal region, we integrated existing global and regional land cover datasets using a workflow including machine learning techniques. This approach seamlessly combines diverse data sources, enhancing representation and accuracy. The resulting map represents high-latitude land cover types at a 1 km spatial resolution, better capturing the spatial heterogeneity of the landscape compared to coarser resolution land surface products, with a total of 35 land cover classes, including 20 forest types (e.g., Larch, Birch, Mixed forests), 6 shrubland classes, and wetlands subdivided into bog, fen, and marsh. To achieve this, we used a global land cover map, the European Space Agency Climate Change Initiative Land Cover data (ESA CCI-LC), as the base map and integrated regional maps across the circumpolar region with finer-resolution land cover information to capture the diversity of land cover types. This approach ensured consistent classification across geopolitical boundaries, while incorporating representative vegetation communities at a region-specific level. Here we documented a workflow used to produce a harmonized circumpolar land cover dataset at 1 km² resolution, encompassing the time period 2000–2023. The hybrid land cover is an open-source product https://doi.org/10.5281/zenodo.17968808 (Briones et al 2025).